diff mbox series

xfs_repair: invalidate dirhash entry when junking dirent

Message ID 20210727191611.GI559212@magnolia (mailing list archive)
State Superseded
Headers show
Series xfs_repair: invalidate dirhash entry when junking dirent | expand

Commit Message

Darrick J. Wong July 27, 2021, 7:16 p.m. UTC 
From: Darrick J. Wong <djwong@kernel.org>

In longform_dir2_entry_check_data, we add the directory entries we find
to the incore dirent hash table after we've validated the name but
before we're totally done checking the entry.  This sequence is
necessary to detect all duplicated names in the directory.

Unfortunately, if we later decide to junk the ondisk dirent, we neglect
to mark the dirhash entry, so if the directory gets rebuilt, it will get
rebuilt with the entry that we rejected.

Signed-off-by: Darrick J. Wong <djwong@kernel.org>
---
 repair/phase6.c |   18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

Comments

Dave Chinner July 27, 2021, 10:17 p.m. UTC | #1 
On Tue, Jul 27, 2021 at 12:16:11PM -0700, Darrick J. Wong wrote:
> From: Darrick J. Wong <djwong@kernel.org>
> 
> In longform_dir2_entry_check_data, we add the directory entries we find
> to the incore dirent hash table after we've validated the name but
> before we're totally done checking the entry.  This sequence is
> necessary to detect all duplicated names in the directory.
> 
> Unfortunately, if we later decide to junk the ondisk dirent, we neglect
> to mark the dirhash entry, so if the directory gets rebuilt, it will get
> rebuilt with the entry that we rejected.
> 
> Signed-off-by: Darrick J. Wong <djwong@kernel.org>

Makes sense.

Reviewed-by: Dave Chinner <dchinner@redhat.com>
diff mbox series

Patch

diff --git a/repair/phase6.c b/repair/phase6.c
index 29259b38..84092269 100644
--- a/repair/phase6.c
+++ b/repair/phase6.c
@@ -237,6 +237,21 @@  dir_hash_add(
 	return !dup;
 }
 
+/* Mark an existing directory hashtable entry as junk. */
+static void
+dir_hash_junkit(
+	struct dir_hash_tab	*hashtab,
+	xfs_dir2_dataptr_t	addr)
+{
+	struct dir_hash_ent	*p;
+
+	p = radix_tree_lookup(&hashtab->byaddr, addr);
+	assert(p != NULL);
+
+	p->junkit = 1;
+	p->namebuf[0] = '/';
+}
+
 static int
 dir_hash_check(
 	struct dir_hash_tab	*hashtab,
@@ -2009,6 +2024,7 @@  longform_dir2_entry_check_data(
 				if (entry_junked(
 	_("entry \"%s\" (ino %" PRIu64 ") in dir %" PRIu64 " is not in the the first block"), fname,
 						inum, ip->i_ino)) {
+					dir_hash_junkit(hashtab, addr);
 					dep->name[0] = '/';
 					libxfs_dir2_data_log_entry(&da, bp, dep);
 				}
@@ -2036,6 +2052,7 @@  longform_dir2_entry_check_data(
 				if (entry_junked(
 	_("entry \"%s\" in dir %" PRIu64 " is not the first entry"),
 						fname, inum, ip->i_ino)) {
+					dir_hash_junkit(hashtab, addr);
 					dep->name[0] = '/';
 					libxfs_dir2_data_log_entry(&da, bp, dep);
 				}
@@ -2124,6 +2141,7 @@  _("entry \"%s\" in dir inode %" PRIu64 " inconsistent with .. value (%" PRIu64 "
 				orphanage_ino = 0;
 			nbad++;
 			if (!no_modify)  {
+				dir_hash_junkit(hashtab, addr);
 				dep->name[0] = '/';
 				libxfs_dir2_data_log_entry(&da, bp, dep);
 				if (verbose)